Amp — Research Preview
@AmpCode
An agentic coding tool, in research preview from @sourcegraph
New @AmpCode feature update: Amp Tab now works in Jupyter Notebooks!
Amp is still up. It falls back to (equivalent) Claude on GCP Vertex when Anthropic's API is down. We care deeply about making Amp the very best agentic coding tool, and we're really happy that most Amp users never noticed any downtime.
Three of us (and customers) noticed basically at the same time that Claude was behaving weirdly: stops mid-sentence, doesn't return anything, feels "lazy". 20min later we're rolling out a switch to GCP Vertex and we're back. Here comes the tokens!
We're experiencing an elevated level of API errors and are currently looking into the issue. ampcodestatus.com
While @AmpCode is great for large, agentic coding sessions… Sometimes you need to make surgical edits to the code yourself. We’ve enabled more fine grained suggestions in Amp Tab, so you can now Tab through a refactor and view each change carefully
I've been testing @AmpCode inside cursor instead of using cursor's agent. I'm very impressed so far. Calling on the Oracle is a sweet feature - it invokes o3 when you need extra intelligence.
Been using @AmpCode for a few weeks now, and the workflow I've settled on is working well. My approach: - Architecture files (bumpers/rails) - Oracle (o3) consultations (large changes, code review) - Proper git flow (feature branches, ez to rollback) - Progress tracking…
We just opened a design engineer position on @AmpCode! ampcode.com/hiring/design-…
Improved JetBrains support just landed in Amp CLI npx @sourcegraph/amp@latest --jetbrains
Tried @AmpCode. Super focused, scarily accurate, sees the entire big picture effortlessly. But something psychological scares me and a bit disapoint me . I’m used to yelling, screaming, and getting angry at agents like Claude code , Windsurf, and Cursor. ( using @WisprFlow)…
We switched a model in Amp and what happened next shocked us
4x tasks merged in 20 mins with @ampcode and vibe-kanban
I've been running five @AmpCode coding agent instances for a couple months. One little piggy works on Emacs. One little piggy builds Node. One little piggy ports old tests. One little piggy reviews code. And one little piggy does wee wee wee random crap the other piggies can't be…
Asked Amp to implement a factorial operator in Boa, a JavaScript engine written in Rust. It ran for ~18min (2 of those waiting for me) and... did it. This is at 16x speed.
In @AmpCode, you can hover over the tools to see their complete descriptions. Super useful examples of how to write effective tool descriptions for AI agents.
We have an outage at the moment related to GCP being down. ampcodestatus.com/incidents/j2h8…